Image Registration Flashcards
What can the meta information of a image contain?
Scale, Orientation, Position…
What is a rigid transformation and how many degree of freedoms does it have?
Rotation + translation, 3/6
What is a similarity transformation and how many degree of freedoms does it have?
rigid + uniform scaling, 4/7
What is a afine transformation and how many degree of freedoms does it have?
rigid + non-uniform scaling + shear, 6/12
How can we perform a shear transformation?
HSH^-1 where H is rotation and S is scaling
What uses has image registration?
Panorama stitching, optical flow, comparission (e.g. medical)…
What is the assumption behind SSD (sum of squared differences) and SAD (sum of absolute differences) cost functions for intensity registration?
- Identity relation between intesity distributions
2. Mono modal
What is the assumption behind CC (Correlation coefficient) cost functions for intensity registration?
- Linear relation between intesity distributions
2. Mono modal.
How can we create a cost function using 2d histograms of intensities for intensity registration?
Search for a registration with where the 2d histogram has most high-value bins. The most used cost function is:
D_NMI = - NMI = (H(I) + H(J)) / H(I,J)
What is the assumption behind D_NMI (Dissimilarity Normal Mutual Information) cost functions for intensity registration?
- Statistical relation
2. Can be multi modal
Where in two images is the disimilarity meassured
Only in the overlap of the registration
How can we leviate the problem of local minimas in Image registration?
Do registration on pyramids of gaussians.
What are some heuristics for image registration initialization?
- Initialize to center of images
2. Inaitalize to center of mass
What kind of search do we usually use for image registration?
Iterative search. (Gradient descent, bayesian optimization…)
What kind of architectures do we often use for image registration with neural networks?
Convolutional similar to architectures used for image segmentation.
What kind of datasets can be used for training of neural networks in image registration?
Syntetic data, (Computer generated video, Flying chairs…)
What is the FlowNet 2.0?
A network for optical flow using a 3 network pipeline for large displacement and a parallel network for small displacement.
What is the idea behind a spacial transformer network?
Add a network that learns parameters for a spacial transform before a classification network. The spacial transform is performed to the input of the classification network. The transform network is not trained directly, but trained by backpropagating classification errors.
What is the idea behind the TeTrIs network?
Feed the image and a shape prior to a neural network that finds the transformation parameters. Perform the transformation on the prior.
What is the formula for CC (Correlation coefficient)
covariance / (sigma_1 * sigma_2)